Greg,
I totally agree with you in your first paragraph here.
Folks earlier on
this thread were suggesting that
implementing a web service portlet
using standard web
app tools would not be hard, and I was trying to
point
out that it would be quite hard to leverage web app features like
caching and session management without a lot of work. I
think you hit
the nail on the head when you say that
"it's not clear how a SOAP
message should be
reconstituted in a normal and repeatable way into
something that the servlets treat as an HTTP request without a
standardized definition". Thanks for being clearer and
much more
concise than I was.
If this hypothesis is true, and portlet development will not
work like
web app development, then I think we need to
be honest about the fact
that we are discarding a lot
of developer knowledge, training, and
experience by
using a SOAP based approach. We are also creating a large
barrier to entry for the large existing codebase of web apps
out there.
I don't mean to imply that we would be able
to create a spec that will
support all existing web
apps if we use HTTP/HTML, but the smaller the
barrier,
the better.
To answer your question, Plumtree's current architecture does
use
HTTP/HTML, and we do use standard proxy caches in
between the portal and
portlet. Since we didn't
define a new caching scheme, portlet
developers can
use standard HTTP caching directives in JSP, ASP, and
Perl, and they have a good strong specification (RFC 2616) as to
what
effect those caching directives will have.
I understand your point about action handling, although I
would argue
again that this is discarding the standard
notions that developers have
about (X)HTML
links. Also, I feel like relative links might be a little
tricky, but I haven't honestly thought that completely
through.
As for your point about session management, I'm not quite sure
I
understand the sentence "Session could be
eliminated as a problem in Java if the portlet container is
in fact a
container rather than a web
application." Can you expand on this point?
Thanks,
Sasha.
-----Original Message-----
From:
PAVLIK,GREGORY (HP-NewJersey,ex2) [mailto:gregory_pavlik@hp.com]
Sent: Wednesday, April 24, 2002 4:49 AM
To: Sasha Aickin; Thomas Schaeck
Cc: Eilon
Reshef; PAVLIK,GREGORY (HP-NewJersey,ex2); Takao Mohri; WSRP
Mailing List (E-mail)
Subject: RE:
[wsrp][markup] First conference call (Summary)
It's not necessarily an easy problem to call from SOAP to
servlets. Axis
may
do what
Thomas states, but that's no guarantee that it's a ubiquitous
capability of all SOAP servers. In addition, even if we assume the
SOAP
request is sent over HTTP and the server gives
the programmer access to
the
RequestDispatcher, it's not clear how a SOAP message should be
reconstituted
in a normal and
repeatable way into something that the servlets treat as
an
HTTP request without a standardized
definition, certainly not without
jeopardizing both
server-side implementation portability and alot of
repetitive coding.
This kind of implementation can be done, but it's not a good
mechanism
for
implementing
portlets. To make it simple and transparent, there really
needs
to be a portlet container unless portlets
are just web applications as
Sasha
seems to be suggesting.
Sasha, what you are describing is essentially the way
Plumtree's portal
architecture is built? Do people
use, for example, HTTP proxy caching
agents
between portlets and portals routinely? I don't really see a
problem if
the
portal
framework has different or new cache management rules for WSRP as
this is an implementation detail of the framework.
With respect to action handling, this should be the
responsibility of a
portlet container, which in Java
is already being defined. Again, I
don't
see a huge problem: hand the action URL from the SOAP handler
to the
container and return the result. It's only
becomes a problem if you take
the
position that WSRP servlets should be implemented as regular web
apps
based
on servlets.
I agree that session is a problem, since we want it to be
replicated or
fault tolerant in the same way we expect
servlet sessions to be (though
persistent state
strikes me as being no different). Session could be
eliminated as a problem in Java if the portlet container is in fact
a
container rather than a web application.
Greg
-----Original Message-----
From: Sasha
Aickin [mailto:AlexanderA@plumtree.com]
Sent: Tuesday, April 23, 2002 11:36 PM
To: Thomas Schaeck
Cc: Eilon Reshef;
PAVLIK,GREGORY (HP-NewJersey,ex2); Takao Mohri; WSRP
Mailing List (E-mail)
Subject: RE:
[wsrp][markup] First conference call (Summary)
Thomas,
I realize that it is possible to implement WSRP services on
Java
application servers in the way that you suggest,
but it seems to me that
a lot of web application
programming paradigms will be so different in
that
scenario that developers will not be able to leverage their web app
development experience, tools, and knowledge.
One example is state and session management. Currently,
servlets depend
on the app server's native session
support, which is usually implemented
with cookies (or
URL rewriting). Several people on our TC have
suggested using a token in the WSRP API as a way to manage states
and
sessions instead of cookies. When we use the
paradigm you suggest (SOAP
code dispatching out to a
servlet), which notion of state/session
management
would we expose to the servlet? If we use the app server's
native cookie-based session management, then the servlet
needs to have
knowledge about how to extract the WSRP
state/session token. If,
instead, we expose the
WSRP session management to the servlet, then we
will
have changed the semantics of the way the servlet works, and
someone may have to write a whole new session mechanism which uses
WSRP
tokens as their key.
Caching is another thing that we are in danger of
rewriting. Standard
HTTP/HTML web apps have
fairly well developed ideas of how caching
should be
done, and those mechanisms have proved remarkably successful
in scaling up the web. However, HTTP caching is not generally
used in
SOAP calls for many reasons, chief among them
being that it is not clear
when two SOAP requests
correspond to the same call into a Web Service
(whereas HTTP can specify this with the Vary header). This
suggests to
me that either WSRP will not support
HTTP-type caching capabilities (the
kind of features
exposed by the Expires, Last-Mod, Etag, and Vary
headers) and will therefore be less feature rich, or we will spend
quite
a while spec'ing out caching rules in our
protocol. If we do spec out
caching rules in our
protocol, we would then need to build
infrastructure
so that web apps written in servlets could easily exploit
those caching mechanisms. We would also have to implement caches
that
understood our new type of SOAP based cache rules
(and potentially even
try to convince proxy server
makers to do the same). This is a
tremendous
amount of work to just get to the level of functionality we
already have with HTTP/HTML web apps.
Action handling is yet another problem that we are looking to
rewrite.
With an HTTP/HTML web app, user actions are
generally handled by links
and forms. However,
since a web service will generally always be at the
same URL, links in the markup will need to be transformed according
to
some scheme to become arguments to the web
service. I could see your
SOAP to servlet scheme
taking in an URL argument to the web service and
dispatching to a servlet which corresponds to the URL, but this would
be
both cumbersome to implement and difficult to make
work with relative
URLs. We are looking at
adding several levels of transformation and
indirection to standard HTML/HTTP links, and I don't see that
action
handling in web services gains any interesting
functionality above and
beyond links in
HTML/HTTP.
My basic point is that, by using SOAP (and decoupling markup
from HTTP),
we are losing many solutions to common
problems that are standard in the
HTTP/HTML
application world, and I am not clear that we are gaining that
much in terms of functionality. By porting concepts
like session
management, caching, and action handling
to web services, we will need
to build and test new
development tools, provide documentation for
development best practices, and train developers in our new solutions
to
these problems. This seems to me like a major
problem for our spec, and
one which will potentially
be a large barrier to entry for developers.
What do people think about this?
Cheers,
Sasha.
-----Original Message-----
From:
Thomas Schaeck [mailto:SCHAECK@de.ibm.com]
Sent: Saturday, April 13, 2002 12:53 PM
To: Sasha Aickin
Cc: Eilon Reshef;
PAVLIK,GREGORY (HP-NewJersey,ex2); Takao Mohri; WSRP
Mailing List (E-mail)
Subject: RE:
[wsrp][markup] First conference call (Summary)
Just as a comment - it will be quite easy to implement WSRP
services as
servlets running on a normal application
server, all one has to do is to
call from the SOAP
code (e.g. SOAP4J or AXIS) into the servlet container
using the request dispatcher that can be obtained within a provider
for
the
SOAP code. Similar
things are probably possible on .NET. So when
starting
from scratch, developing WSRP services
will be about as simple or hard
as
developing web apps.
But of course it is true that existing web apps need to be
changed to
run
as WSRP
services...
Best regards,
Thomas
Sasha Aickin <AlexanderA@plumtree.com> on 04/12/2002
05:48:27 AM
Please respond to Sasha Aickin
<AlexanderA@plumtree.com>
To: Eilon Reshef
<eilon.reshef@webcollage.com>, "PAVLIK,GREGORY
(HP-NewJersey,ex2)"
<gregory_pavlik@hp.com>, Thomas
Schaeck/Germany/IBM@IBMDE, Takao
Mohri
<mohri.takao@jp.fujitsu.com>
cc: "WSRP Mailing List (E-mail)"
<wsrp@lists.oasis-open.org>
Subject: RE: [wsrp][markup] First conference call
(Summary)
Eilon,
While this is true (and I'm generally a big fan of
integrating with
existing systems), supporting
HTML doesn't seem like the big barrier to
supporting
JSP and Perl scripts. The big problem with that, it seems,
is
that we've committed
ourselves to the SOAP web services world.
Rewriting
a web app to be a web
service is decidely non-trivial; you can lose out
on
a lot of things that J2EE server and
IIS give you for free (session
management being a huge
one), and fitting in with the web service
programming paradigms is very different than writing a web app
page.
As a sidenote, I have to admit that sometimes I get
frustrated that we
are
trying
to solve a lot of problems that have already been solved with
standard web app development tools. Session
management,
authentication,
privacy, efficient caching, and other issues are all fairly
well
understood using the basic HTML/HTTP
solution, and I feel like we're
rewriting a lot
of that. Not only is it a lot of work, it will be hard
for us to document and for developers to learn.
Does anyone else
sometimes feel this way, or am
I the only one?
That all being said, I think that we have to support
HTML if only for
the
huge
amount of developer knowledge of HTML (vs. XHTML).
Cheers,
Sasha.
-----Original Message-----
From: Eilon
Reshef [mailto:eilon.reshef@webcollage.com]
Sent: Wednesday, April 10, 2002 8:58 AM
To: 'PAVLIK,GREGORY (HP-NewJersey,ex2)'; 'Thomas
Schaeck'; 'Takao
Mohri'
Cc: 'WSRP Mailing List (E-mail)'
Subject:
RE: [wsrp][markup] First conference call (Summary)
I do believe we have to offer a migration path for
people that have or
even prefer plain old HTML -
expecting everybody to go back to their
Perl
scripts and JSP templates and rework them
might provide a barrier to
entry
that we can avoid by simply supporting HTML.
-----Original Message-----
From:
PAVLIK,GREGORY (HP-NewJersey,ex2) [mailto:gregory_pavlik@hp.com]
Sent: Wednesday, April 10, 2002 11:48 AM
To: 'Thomas Schaeck'; Takao Mohri
Cc: PAVLIK,GREGORY (HP-NewJersey,ex2); WSRP Mailing List
(E-mail)
Subject: RE: [wsrp][markup] First conference
call (Summary)
Is there a general statement on whether the markup
subcommittee intends
to
treat
HTML and XHTML interchangeably? My initial thought was that WSRP
would
do better to standardize on
XHTML, which would have a number of
advantages
including the ability of the consumer to treat the
fragment as an XML
fragment. It's not terribly
difficult to tidy up a piece of HTML and
repurpose it
as well-formed/valid XHTML when dealing with legacy
content.
An
early focus
on XHTML might position us better for dealing with devices
in
the long run, as it appears
that WML and cHTML are going to be
secondary
targets for the TC.
Greg
-----Original Message-----
From:
Thomas Schaeck [mailto:SCHAECK@de.ibm.com]
Sent: Tuesday, April 09, 2002 5:43 PM
To: Takao Mohri
Cc: PAVLIK,GREGORY
(HP-NewJersey,ex2); WSRP Mailing List (E-mail)
Subject: Re: [wsrp][markup] First conference call (Summary)
There is not something like "the markup language of
WSRP", the WSRP
protocol needs not to be tied to
any particular markup language, the
protocol and
the markup exchanged can be viewed as orthogonal.
However, WSRP will need to define rules for valid markup
fragments and
common styles with the
priorities
1. (X)HTML
2.
WML, cHTML, VoiceXML...
It would be interesting to see the rules that would
define valid XHTML
Basic fragments.
Takao, could you post an XHTML Basic example document to
the mailing
list
and an
outline of fragment rules for XHTML Basic ? Which tags would be
allowed in aggregatable XHTML Basic fragments and which
tags would not
be
allowed to
use in fragments ?
Best regards,
Thomas
Takao Mohri <mohri.takao@jp.fujitsu.com> on
04/09/2002 01:26:58 PM
Please respond to Takao Mohri
<mohri.takao@jp.fujitsu.com>
To: "PAVLIK,GREGORY
(HP-NewJersey,ex2)" <gregory_pavlik@hp.com>
cc: "WSRP Mailing List (E-mail)"
<wsrp@lists.oasis-open.org>
Subject: Re: [wsrp][markup] First conference
call (Summary)
Greg,
Yes, as you mentioned, XHTML Basic is becoming a
standard for cellular
phones.
In WAP 2.0, XHTML Mobile Profile (an extention of XHTML Basic) is
a
standard markup language. In Japan, cellular phones
that support WAP
2.0
are
already on sale by KDDI. NTT Docomo also has announced support of
WAP
2.0.
In
the near future, WAP 2.0, i.e, XTHML Basic will be used by many
mobile
devices.
If the markup language of WSRP is based on XHTML
Basic,
it makes support of mobile devices much
easier.
Regards,
Takao
"PAVLIK,GREGORY (HP-NewJersey,ex2)" wrote:
>
> Gino,
>
> Could you clarify if your intent
is to consider XHMTL/XHTML Basic as
a
part
> of the HTML focus? This
wasn't clear from the notes. There appears to
be
a
> broad
convergence in the mobile space, both with respect to WML and
NTT
> Docomo, to standardizing on XHTML
Basic + CSS. XHTML is going to be
here
on
> the device very soon, perhaps
quicker than on the desktop, making
this
a
> relatively high priority item. Fujitsu
may have some comments on this
as
> well.
>
> Greg
>
>
-----Original Message-----
> From: Gino
Filicetti [mailto:gfilicetti@bowstreet.com]
> Sent: Friday, April 05, 2002 3:22 PM
> To: WSRP Mailing List (E-mail)
>
Subject: [wsrp][markup] First conference call (Summary)
>
> Team,
>
> Attached are the meeting notes
from today's Markup Subcommittee
meeting.
> Comments and suggestions are welcome.
>
> One thing I did notice was
that there were quite a few participants
today
> that aren't on the actual list
of official members for this team.
Since
this
> team is quite small, I
think it'd be great if we could get some of
you
to
>
officially join and help out with the effort. Unfortunately, I wasn't
able
> to capture the names
of ALL the attendees this time around (I'll be
sure
to
> record the
con-call next time), but please send me an email so we can
add
> your name to official
list and flesh out the team a little more.
>
> Currently, the official list is:
>
> Gino Filicetti
> David Taieb
> Khurram Mahmood
> Susan Levine
> Michael
Hillerman
>
>
Thanks,
>
> . . . . . .
. . . . . . . . . . . . . . . . . .
> Gino
Filicetti | Software Engineer
> One Harbour
Place, Portsmouth, NH 03801
> T 603.559.1692
| gfilicetti@bowstreet.com
> w w w . b o w s
t r e e t . c o m
>
>
----------------------------------------------------------------
> To subscribe or unsubscribe from this elist use
the subscription
> manager: <http://lists.oasis-open.org/ob/adm.pl>
// Takao MOHRI (mohri.takao@jp.fujitsu.com)
// FUJITSU LABORATORIES LTD.
----------------------------------------------------------------
To subscribe or unsubscribe from this elist use the
subscription
manager: <http://lists.oasis-open.org/ob/adm.pl>
----------------------------------------------------------------
To subscribe or unsubscribe from this elist use the
subscription
manager: <http://lists.oasis-open.org/ob/adm.pl>
----------------------------------------------------------------
To subscribe or unsubscribe from this elist use the
subscription
manager: <http://lists.oasis-open.org/ob/adm.pl>